Publication
PORTAL - Plant breeding using robotics and AI for advanced data analysis and decision-making in virtual space
Christoph Tieben; Benjamin Kisliuk; Muhammad Moiz Sakha; Naeem Iqbal; Florian Daiber; Matthias Enders; Antonio Krüger; Joachim Hertzberg
44. GIL-Jahrestagung, Gesellschaft für Informatik in der Land-, Forst- und Ernährungswirtschaft (GIL-2024), 2/2024.
Abstract
In the process of plant breeding for developing robust and high-performing field crops, it is unavoidable to regularly evaluate a variety of candidates in plot trials. A large number of parameters and characteristics are accessed at the various stages of plant development. Data collection and evaluation can only be carried out by appropriately trained experts and must take place in special stages at a high frequency in different locations.
An autonomous robotic approach can significantly reduce the effort of breeding and open up new possibilities. Continuous monitoring of the individual plots through regular, highly accurate, acquisition of multi-modal data provides the basis for creating a detailed, three-dimensional model of the breeding plots. This model will be displayed for the breeders within a virtual reality (VR) environment. This will enable them to assess a virtual breeding nursery. In this virtual nursery, not only data in the range of the light spectrum visible to humans will be presented, but also false-color views of data in the UV, NIR, and IR range will be available.
The process of trait recording can thus be supported by additional information. Data from previous breeding periods can be directly compared with current data. This provides a completely new basis for decision-making in plant breeding. The collected data can also be used to train classification models to identify potential diseases or particularly desirable traits at an early stage. In combination with the close-meshed data collection by an autonomously acting robot, new possibilities will arise for the targeted selection of future varieties.